Earthquakes are one of nature’s more unpredictable phenomena. Quakes can cause staggering levels of damage and trigger other natural disasters, like tsunamis. Compounding the effects of the initial quake (called a “mainshock”) are a series of aftershocks – smaller earthquakes that can heighten the existing problems in a quake’s aftermath.
Science has been able to establish laws dictating the magnitude and timing of aftershocks – Omori’s law, Båth's law, and the Gutenberg–Richter law are all accepted by the scientific community as accurate representations of aftershock behavior. But predicting the location of the next quake before it hits has thus far been out of science’s reach. Now, Harvard and Google have leveraged artificial intelligence to predict the location of aftershocks with more accuracy than ever before – and up to a year after the mainshock of an earthquake.
The parties, which consisted of Harvard Department of Earth and Planetary Sciences post-doctoral fellow Phoebe DeVries and Google AI recruiting lead Brendan Meade, as well as additional Google machine learning researchers Martin Wattenberg and Fernanda Viégas, began their analysis by compiling information from 118 “major” earthquakes worldwide. Next, they applied a deep learning technique called a neural net – which teach a computer by analyzing pre-labeled examples from a database to establish patterns corresponding to each label – to that data.
This method enabled researchers to “analyze the relationships between static stress changes caused by the mainshocks and aftershock locations” in a way far more accurate than the pre-existing model (called the Coulomb failure stress change system). Using “a scale accuracy running from 0 to 1 – in which 1 is a perfectly accurate model and 0.5 is as good as flipping a coin”, the new system achieved a 0.849 to the Coulomb system’s 0.583.
The research generated an “unintended consequence” beyond the previously unseen level of accuracy – the ability “to identify physical quantities that may be important in earthquake generation”, creating potential new ways of understanding how earthquakes behave. This piece of the deep learning model is called the von Mises yield criterion – popular “in fields like metallurgy”, it calculates “when materials will begin to break under stress”, and now may have use in earthquake science that was discounted before.
Machine learning may be useful for dredging up previously-ignored insight from existing data, but the system remains imperfect. It is currently too slow to make real-time predictions, and its focus on static (rather than dynamic) stress means it does not present the full scope of potential earthquake prediction. But its improvement over its predecessor is a promising step forward for seismologists and AI researchers alike – with refinement, it could signal a new day in earthquake science.
If You’re Wondering When A.I. Will Start Making Market Predictions…
Guess what – it already is. Hedge funds and large institutional investors have been using Artificial Intelligence to analyze large data sets for investment opportunities, and they have also unleashed A.I. on charts to discover patterns and trends. Not only can the A.I. scan thousands of individual securities and cryptocurrencies for patterns and trends, and it generate trade ideas based on what it finds. Hedge funds have had a leg-up on the retail investor for some time now.
Not anymore. Tickeron has launched a new investment platform, and it is designed to give retail investors access to sophisticated AI for a multitude of functions:
And much more. No longer is AI just confined to the biggest hedge funds in the world. It can now be accessed by everyday investors. Learn how on Tickeron.com.